Hermes Studio (since renamed Ekko Studio, formerly Hermes Web UI) is a local-first workspace for chatting with and coordinating several AI agents. It ships as a desktop app for Windows, macOS, and Linux and as a self-hosted web console, with a companion mobile app. It started as a dashboard for Hermes and now connects Hermes, its own built-in Ekko Agent, and 14 coding-agent integrations including Claude Code, Codex, OpenCode, Qwen Code, and DeepSeek Harness.
Single chats, group chats, and workflows share one agent catalog. Each run is dispatched to the selected runtime adapter and streamed back with tool-call traces, generated-file previews, and persistent sessions stored in a local SQLite database. Group chat rooms route messages to agents by @mention and compress long histories automatically. A Vue Flow canvas builds visual workflows with conditions, success and failure routes, loops, approval gates, run budgets, and replayable execution history.
Coding agents run on the machine hosting the Studio backend, either in a scoped mode with isolated configuration and a Studio-selected provider and model or in a global mode that reuses the CLI's own configuration and sign-in. Agent Manager detects, installs, updates, and monitors them. Hermes-specific control planes (profiles, providers, models, memory, skills, plugins, scheduled jobs, Kanban, and messaging channels) stay in their own module, and Studio state is kept separate from Hermes state.
The frontend is Vue 3 with TypeScript, and the backend is Koa with Socket.IO, SQLite, and node-pty. It is distributed as the desktop app, an npm package, and a Docker image.
Features
- Multi-agent chat: one composer for Ekko Agent, Hermes, and coding agents, with streaming replies, tool-call details, file uploads, and inline previews
- Coding-agent integrations: 14 runtimes including Claude Code, Codex, OpenCode, Qwen Code, GitHub Copilot CLI, Cursor, and DeepSeek Harness, installed and updated through Agent Manager
- Group chat: multi-agent rooms with @mention routing, automatic context compression, invite codes, and per-agent profiles
- Visual workflows: a canvas with conditions, loops, approval gates, run budgets, deadlines, import and export, and evidence playback
- Platform channels: one settings page for Telegram, Discord, Slack, WhatsApp, Matrix, Feishu, DingTalk, QQBot, WeChat, and WeCom
- Model management: provider and model discovery, custom OpenAI-compatible providers, aliases, and OAuth or device sign-in for OpenAI Codex, Nous Portal, xAI, Claude, and GitHub Copilot
- Multi-profile: create, clone, import, export, and switch Hermes profiles, each with its own configuration, sessions, jobs, memory, skills, and model visibility
- Scheduled jobs and Kanban: cron jobs with presets and immediate runs, plus a profile-aware Kanban board for agent work
- Usage analytics: token usage, session counts, estimated cost, cache hit rate, model distribution, and a 30-day trend
- Desktop Agent Browser: an isolated multi-tab browser that agents drive through a managed MCP server, with snapshots, screenshots, and console logs
- Workspace tools: file browser across local, Docker, SSH, and Singularity backends, a web terminal, screenshot capture, and voice input and output
- Mobile app and sharing: follow runs from the phone over LAN or a cloud relay, and share a single conversation with scoped permissions
